Antibodies are of major interest in the development of anti-cancer therapeutics and as reagents for detection of specific molecules in many areas of biology. While antibodies can be produced by immunization of animals or by phage and yeast display, many epitopes of great importance because of their clinical or biological effects resist current methods of antibody development. We have developed a computational method for antibody design in the Rosetta program based on structural bioinformatics studies we have previously published. In this proposal, we describe three aims: 1) to significantly extend structural bioinformatics studies of antibodies, highlighting detailed sequence-structure relationships, the effects of germline origin on the framework and CDR structures, and an effort to rebuild and refine poorly determine X-ray crystallographic structures of antibodies in the PDB; 2) to perform computational benchmarking to optimize sampling and scoring functions for antibody design using two novel metrics and suitable measures of statistical significance; 3) to test our antibody design methods in the laboratory on several carefully chosen targets. As they mature, these methods are likely to make significant impact on the field of antibody engineering.